edf2csv 0.5.96 → 0.5.98

package/CHANGELOG.md

@@ -3,6 +3,49 @@
 Notable changes to edf2csv. Versions follow [semantic versioning](https://semver.org); while the
 major version is 0, a minor bump may contain breaking changes.
 
+## 0.5.98
+
+### Fixed: "what --info can and can't tell you" was wrong in both directions
+
+cli-reference points readers at that section for the answer, which makes being wrong there
+worse than being silent. It was wrong twice.
+
+It said `--info` "also raises `ANNOTATION_DECODE_FAILED`", unqualified. It raises it for what
+it read, and on a continuous file that is sixteen records — so `two-annotation-channels.edf`
+warns about three unreadable events when converted and says nothing at all under `--info`. Its
+byte-identical discontinuous twin warns either way, because there the whole channel is read.
+
+And its list of what `--info` cannot raise — `NO_ANNOTATIONS`, `STALE_OUTPUT`, the EDF+C
+contradiction — left out the `NO_SAMPLES` that reports a signal file *not written*.
+`annotations-only.edf` raises it on conversion and not under `--info`. The per-channel
+`NO_SAMPLES`, about a channel carrying no samples, comes from the header and is raised, so the
+code alone does not settle it and the page now says which form it means.
+
+Held to it by a sweep rather than by review: every fixture is described and converted, and any
+code the conversion raises that `--info` does not has to be one the page names. That is how the
+two missing ones were found, and it is what would find the next.
+
+## 0.5.97
+
+### Fixed: three pages said `--info` reads only the header, on files where it reads records
+
+"`--info` reads only the header for plain EDF and continuous EDF+, so it returns in
+milliseconds whatever the file's size." Getting-started said it, recipes said it twice.
+
+Since 0.5.46 it reads up to sixteen records' annotation slots of a *continuous* EDF+ as well,
+to find where the recording begins — which is why it raises `ANNOTATION_DECODE_FAILED` on
+`lost-timekeeping.edf`, a continuous file, and prints `Timed from 0.500s` for
+`fractional-start.edf`, another one. Both are behaviours the pages elsewhere describe and
+recommend.
+
+warnings-and-errors has had this right since 0.5.37 — "the first few records of a continuous
+one to find where the recording begins" — and cli-reference points readers there for the
+answer, so three pages contradicted the fourth about how much of a file a command touches.
+
+All three say what it does now, and link to the page that sets out which warnings follow. The
+test asserts the behaviour first and the prose second: if `--info` ever really did stop reading
+records, the pages would be right and this test is what should fail.
+
 ## 0.5.96
 
 ### Fixed: api.md said `readAnnotations` returns three counts, then named four

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edf2csv",
-  "version": "0.5.96",
+  "version": "0.5.98",
   "description": "Convert EDF, EDF+ and BDF biosignal recordings (European Data Format) to CSV from the command line. Local, streaming, and never resamples or alters units.",
   "keywords": [
     "edf",